35 Domains 12 Stacks BM25 + Synonyms Compare Mode Interactive REPL 652 Tests · CI 100% Source Citations Python 3.8+ Zero Dependencies MIT License
// Staff-Engineer-in-a-Box

Backend Pro Max
Design Intelligence

Curated, BM25-searchable backend & distributed-systems intelligence across 35 domains and 12 language stacks — every row sourced, auditable, and freshness-tracked. Drop it into Claude Code, Cursor, Copilot, or any AI assistant.

Get on GitHub How it works ↓ View Domains →
📤 Export to Obsidian · Notion · Org-mode backendpro export --format obsidian
bash — backend-pro-max
$ pip install backendpro
Successfully installed backend-pro-max-skill
$ backendpro "kafka exactly once delivery"
domain: messaging   score: 7.69   confidence: high
do: Use transactional.id + enable.idempotence=true
don't: Rely on at-least-once for financial events
severity: CRITICAL
$ backendpro compare "Kafka" "RabbitMQ" --domain messaging
| Field | Kafka | RabbitMQ |
| Throughput | Very High      | Medium-High    |
| Ordering | Per-partition | FIFO per queue |
| Replay | Yes           | Limited        |
$ backendpro --interactive  # REPL for design sessions
$ backendpro decide "Kafka vs Pulsar vs Pub/Sub" --constraints throughput=high
recommendation: Apache Pulsar   candidates: 5
constraints: throughput:high, cloud:gcp
match: 2/2 ✓
$
35
Domains
12
Stacks
BM25
+ Synonyms
0
Dependencies
652
Tests · CI Green
6
AI Tools Supported

// Knowledge Base

35 Backend Domains

Every domain contains opinionated, source-citable, staff-engineer-grade rows with Do / Don't examples and severity ratings.

pattern
pattern
Saga, CQRS, Outbox, Circuit Breaker, Bulkhead, Idempotency, Leader Election, Rate Limiting, Sharding
22 patterns
database
database
Postgres, MongoDB, Cassandra, DynamoDB, Redis, ClickHouse, vector DBs, S3, etcd
20+ engines
messaging
messaging
Kafka, Redpanda, Pulsar, RabbitMQ, NATS, SQS, SNS, Kinesis, Pub/Sub
12 brokers
cache
cache
In-process LRU, Redis cluster, Memcached, CDN, negative caching, Bloom filters, L1+L2 hybrid
10 strategies
cloud
cloud
AWS / GCP / Azure / Cloudflare service mapping and equivalents across all major primitives
3 providers
observability
observability
Prometheus, Grafana, Loki, Tempo, OpenTelemetry, Pyroscope, SLO frameworks, Datadog
15+ tools
security
security
OWASP Top 10, CSRF, SSRF, supply chain (SLSA, Sigstore), zero-trust, TLS, PII, SAST
Critical severity
auth
auth
OAuth 2.0+PKCE, OIDC, JWT, SAML, mTLS, passkeys, RBAC/ABAC/ReBAC, SCIM, workload identity
12 mechanisms
api
api
REST, GraphQL, gRPC, WebSocket, SSE, HTTP/2+3, Webhooks, JSON-RPC
10 protocols
container
container
Docker, Kubernetes, EKS/GKE/AKS, Helm, ArgoCD, Istio, Envoy, Karpenter
20+ tools
consistency
consistency
Linearizability, CAP, PACELC, Raft, Paxos, 2PC, snapshot isolation, Lamport clocks
8 models
reliability
reliability
SLO/SLI/error budget, timeouts, retries, circuit breaker, graceful shutdown, chaos engineering
18 practices
scaling
scaling
Autoscaling (HPA/KEDA), sharding, backpressure, hedged requests, geo-distribution, CDN
16 techniques
performance
performance
N+1, GC pauses, hot keys, tail latency, thundering herd, cold starts, TLS overhead
14 anti-patterns
testing
testing
Unit, integration (Testcontainers), contract (Pact), fuzz, mutation, chaos, load, smoke
10 types
architecture
architecture
Monolith, microservices, serverless, event-driven, hexagonal, DDD, CQRS+ES, cell-based
12 styles
cicd
cicd
GitHub Actions, GitLab CI, ArgoCD, Flux, Argo Rollouts, Renovate, SonarQube, GHAS
15+ tools
iac
iac
Terraform/OpenTofu, Pulumi, AWS CDK, CloudFormation, Bicep, Ansible, Crossplane, Helm
9 tools
data
data
Spark, Flink, Kafka Streams, dbt, Airflow, Iceberg/Delta/Hudi, Debezium, vector DBs
20+ tools
language
language
Go, Java, Kotlin, Python, Rust, Node.js/TS, C#, Scala, Elixir, Ruby, PHP, C++
12 languages
antipattern
antipattern
Distributed Monolith, God Service, Dual Writes, Sync-over-Async, Chatty Microservices, Unbounded Retry, N+1 Query
15 anti-patterns
cost
cost
FinOps, spot instances, reserved capacity, right-sizing, cost allocation tags, waste detection
12 strategies
mobile-backend
mobile-backend
BFF pattern, push notifications, offline sync, GraphQL federation, edge caching, app config
10 patterns
multi-tenant
multi-tenant
Schema-per-tenant, row-level isolation, silo vs pool, noisy neighbor, tenant routing
8 models
api-contract
api-contract
OpenAPI, AsyncAPI, Protobuf, consumer-driven contracts (Pact), schema registry, versioning
8 standards
data-engineering
data-engineering
ELT/ETL, data lakehouse, medallion architecture, schema evolution, data contracts, lineage
14 practices
edge
edge
Edge compute (Cloudflare Workers, Lambda@Edge), CDN logic, geo-routing, edge databases
9 patterns
incident
incident
On-call, runbooks, blameless postmortems, incident severity, SLA communication, war rooms
11 practices
interview
interview
System design interview patterns, capacity estimation, trade-off analysis, communication frameworks
20 topics
capacity
capacity
QPS estimation, storage projections, bandwidth math, replication factors, back-of-envelope
10 formulas
compliance
compliance
SOC 2, GDPR, HIPAA, PCI-DSS, ISO 27001, data residency, audit logging, retention policies
8 frameworks
latency-numbers
latency-numbers
L1/L2 cache, RAM, SSD, network hops, cross-region, DNS, TLS handshake — Jeff Dean's numbers updated
25+ benchmarks
migration
migration
Strangler Fig, blue-green DB, dual writes, shadow traffic, feature flags, zero-downtime schema
12 strategies
ml-platform
ml-platform
Feature stores, model serving, A/B testing, GPU scheduling, MLOps pipelines, vector search
14 components
release
release
Canary, blue-green, rolling, feature flags, dark launches, ring deployments, rollback strategies
10 strategies

// Stack Guidelines

12 Language Stacks

What would a staff engineer say in code review? Opinionated Do / Don't with real code examples, per stack.

Go
Go
context.Context, errgroup, pgx/sqlc
concurrencyhttptesting
Java
Java / Spring
Virtual threads (Loom), constructor DI
loomflywaynative
Python
Python / FastAPI
async-all-the-way, Pydantic v2, uv
asyncruffstructlog
Node.js
Node / Express
Fastify > Express, zod, pino, Vitest
fastifyotelundici
Rust
Rust / Axum
Tokio + Tower, sqlx compile-time queries
tokiotracingthiserror
C#
C# / ASP.NET
Minimal APIs, Polly v8, Native AOT
minimal-apief-coreaot
Kotlin
Kotlin / Spring
Coroutines, structured concurrency, kotest
coroutinesjooqkotest
Scala
Scala / Akka
Pekko, Typed actors, Cats Effect / ZIO
pekkostreamszio
Elixir
Elixir / Phoenix
OTP, GenServer, Broadway, libcluster
otpliveviewbroadway
Ruby
Ruby / Rails
Modular Rails, Sidekiq, Solid Queue
packwerkpumabullet
PHP
PHP / Laravel
Octane (Swoole/FrankenPHP), PHPStan
octanehorizonopcode
C++
C++
C++20, RAII, jthread, sanitizers, CMake
c++20conanclang-tidy

// Beyond search

Built for staff-level decisions

Search is the floor. Compare modes, freshness audits, an interactive REPL, and confidence scores turn the skill into a real decision-support tool.

compare
compare mode
Side-by-side markdown table of two or more options — perfect for ADRs.

backendpro compare
"Kafka" "RabbitMQ"
--domain messaging
ADR-ready output
repl
interactive repl
Exploratory design sessions with /d, /s, /all, /cmp, /stale.

backendpro --interactive
Design-session friendly
freshness
freshness tracking
Every row can carry a Last Updated date. Filter or audit.

backendpro --stale
--domain pattern
--max-age-months 18
Quarterly audit ready
confidence
confidence scores
Every result includes a BM25 score and a high / medium / low label so the AI agent knows when to trust the hit.

--min-score 1.5
Agent-grade signal
synonym
synonym hybrid
Plain-English queries reach the right rows even when keywords don't overlap. "Partial failure" → Saga.

--no-expand to disable
Bridges keyword gaps
cached
cached index
Lazy in-memory BM25 index, mtime-invalidated. Sub-millisecond repeat queries — agent loops stay snappy.
No recompute per query
validator
csv validator
Schema-validates every CSV against its declared columns and date formats. Wired into CI to block bad rows.

backendpro-validate
Enforced on every PR
tested
tested + linted
652 pytest cases covering ranking quality, golden queries (239 assertions), edge cases, decision intelligence, intent classification, trust/verifiability, freshness, provenance, conflict detection, dedup, and coverage analysis. Ruff lint + CSV validation on Python 3.9 / 3.11 / 3.12.
Green on every push
intent
intent classifier
Auto-detects query intent — comparison, troubleshoot, migration, incident, definition, best-practice, checklist — and applies structured templates to output.

backendpro "how to fix
connection pool exhaustion"
--intent troubleshoot
7 intent types
hybrid
hybrid retrieval
Optional embedding-based search via sentence-transformers + Reciprocal Rank Fusion with BM25. Graceful fallback to pure BM25.

pip install backendpro[semantic]
backendpro "conceptual query"
--engine hybrid
BM25 + embeddings
rerank
cross-encoder rerank
Optional cross-encoder re-ranking for precision-critical queries. Graceful fallback when not installed.

pip install backendpro[rerank]
backendpro "saga pattern"
--rerank
Precision boost
intent
decide
Constraint-aware ranked recommendations. Extracts facets (throughput, latency, cloud, consistency) and scores every candidate.

backendpro decide
"Kafka vs Pulsar vs Pub/Sub"
Beats plain LLM comparison
adr
adr generator
Auto-generates Architecture Decision Records in Markdown with context, options, trade-offs, and citations.

backendpro adr
"Redis vs Memcached"
Markdown ADR output
design
design doc
Capacity-aware design documents with QPS estimates, storage projections, and replication calculations.

backendpro design
"Postgres for 50M DAU"
Capacity math included
citations
source citations
Every row carries a Source URL, Source Type, and Last Updated date. 100% coverage across 685 rows — official docs, RFCs, papers, OWASP cheat sheets.

backendpro-validate --strict
685/685 rows sourced
conflict
conflict detector
12 curated architectural tension rules surface trade-offs: retry vs latency, cache vs consistency, outbox vs CDC, sharding vs joins, and more.

backendpro conflicts
--domain cache
Known trade-offs surfaced
freshness
auto-freshness
Weekly GitHub Action flags stale rows (>18 months) and broken Source URLs. Opens issues per domain — no manual auditing needed.

backendpro-validate
--check-urls
Weekly automated audit
tested
golden queries
239 retrieval assertions across all 35 domains, 12 stacks, cross-domain, and compare mode. Guards BM25 ranking quality on every PR — actionable failure output shows expected vs actual top-5 with scores.
239 assertions · CI gate
dedup
dedup
BM25 pairwise self-similarity detects near-duplicate rows within and across domains. Configurable threshold, allowlist for intentional overlaps.

backendpro dedup
--cross-domain --threshold 0.7
Catches silent drift
coverage
coverage report
Per-domain coverage analysis with gap detection (missing categories), thin detection (<3 rows), and summary stats. Drives contribution by making gaps visible.

backendpro coverage
--domain messaging --json
Gaps made visible

// MCP Integration

Model Context Protocol

Backend Pro Max runs as an MCP server — Claude Desktop, Cursor, Windsurf, and any MCP-compatible agent can query 685 knowledge rows natively.

connect
Zero-config MCP server
Install once, add to claude_desktop_config.json — Claude sees all 35 domains instantly.
search
6 MCP Tools exposed
search, compare, decide, lint, conflicts, learn — all callable by the agent directly.
trust
Cited, verifiable answers
Every response includes source URLs, confidence scores, and freshness dates — no hallucination.
setup
One-line setup
pip install backendpro
backendpro --mcp
Backend Pro Max running in Claude Desktop via MCP

// Workflow

How AI agents use this

From prompt to production-grade, cited advice in 5 intelligent steps. The model searches before answering.

01
Your Prompt
You ask your AI assistant a backend question. Natural language, no special syntax needed.
02
Domain Detection
The skill detects which domain applies — messaging, reliability, pattern, stack — from your query automatically.
03
BM25 + Synonym Expansion
Pure Python BM25 ranks rows across 35 domain CSVs, with synonym expansion ("partial failure" → "compensation, saga") and an mtime-cached in-memory index. Every result carries a confidence score. No network, no model.
04
Cited Answer
The model answers with the right library, code pattern, and a citation to the exact row. Reviewers can verify.
05
Cross-domain Synthesis
Complex queries pull from multiple domains — a saga answer combines pattern + messaging + consistency + reliability.
— user prompt
# Your question to the AI assistant

You:
"Add retries to our outbound HTTP client
without melting the dependency.
We're using Python + FastAPI."

# Any natural language works.
# No special syntax required.
# Auto-detection from query keywords

query_keywords: ["retry", "http", "circuit"]
detected_domains:
reliability — score 0.91
pattern    — score 0.84
stack     — python-fastapi

# Skip --domain and engine picks it.
# Pure Python BM25 — no installs

$ backendpro "retry backoff jitter" \
--domain reliability \
--stack python-fastapi

Searching reliability.csv …
row: retry-exponential-jitter
score: 0.96
severity: HIGH
# Model answers with citation

Answer (reliability › retry-backoff-jitter):

✓ Do: Use tenacity with exponential
backoff + full jitter. Wrap only
outbound calls, not route handlers.

✗ Don't: Retry on 4xx. Cap at
3 attempts with a budget timeout.

lib: tenacity + httpx
# Saga pulls 4 domains at once

$ backendpro "saga distributed tx" --all

Cross-domain synthesis:
pattern    → choreography vs orchestration
messaging  → outbox + idempotent consumers
consistency→ eventual, not 2PC
reliability→ compensating transactions

// Get Started

Drop into any AI tool

Install once with pip, then load the skill into your favourite assistant.

🟣 Claude Code
Adds SKILL.md to your project
pip install backendpro
mkdir -p .claude/skills
backendpro --load claude
🟦 Cursor
Adds .cursor/rules/backend.mdc
pip install backendpro
backendpro --load cursor
🌊 Windsurf
Adds .windsurfrules to your project
pip install backendpro
backendpro --load windsurf
🐙 GitHub Copilot
Adds .github/copilot-instructions.md
pip install backendpro
backendpro --load copilot
🔷 Continue
Adds AGENTS.md to your project
pip install backendpro
backendpro --load continue
⚙️ CLI / MCP
Search & serve from the terminal
pip install backendpro
backendpro "redis cluster" --json
backendpro --list
backendpro-mcp
🧪 Contributors
Tests, lint & CSV validator
pip install -e ".[dev]"

pytest # 652 tests
ruff check src tests # lint
backendpro-validate # CSVs
// Ready to ship?

Stop guessing. Start citing.

Ground your AI assistant in opinionated, staff-engineer-grade backend knowledge. Zero setup, zero dependencies.

Get on GitHub Browse Domains →
Open Source MIT License Zero Dependencies Python 3.8+
Built for engineers who actually ship distributed systems. If this saves you one outage, ⭐ the repo.